Robot performance prediction method, device and equipment and storage medium

Abstract

The embodiment of the invention provides a robot performance prediction method and device, equipment and a storage medium, which are used for realizing accurate prediction of robot performance. Comprising the following steps: simulating the rigid body dynamic model based on a joint driving instruction output by the operation control model, and obtaining mechanical and kinematic response parameter information output by the rigid body dynamic model; consistency comparison is carried out on the mechanics and kinematics response parameter information, corresponding prediction parameter information output by the operation control model and robot actual measurement parameter information, the rigid body dynamic model is calibrated according to a consistency comparison result, and a calibrated model is obtained; a key bearing part of the calibrated model is subjected to flexibility, a rigid-flexible coupling dynamic model is obtained, actually-measured robot motion data are input into the rigid-flexible coupling dynamic model, and a dynamic stress result of the key bearing part is determined; according to the dynamic stress result, durability evaluation is conducted on the key bearing component, and an evaluation result is obtained.

Description

Technical Field

[0001] This disclosure relates to the field of robotics, and in particular to a method, apparatus, device, and storage medium for predicting robot performance. Background Technology

[0002] With the rapid development of robotics technology, robot products are transitioning from experience-based design to predictive and precision design. In particular, the emergence of legged robots such as quadruped robots and humanoid robots has placed higher demands on the structural strength, motion accuracy, reliability, and safety of robots.

[0003] Predicting robot performance through mechanical simulation before manufacturing a physical prototype has become a key technology for shortening R&D cycles, reducing R&D costs, and enhancing product competitiveness.

[0004] However, existing robot mechanics simulation methods suffer from problems in practical applications, such as large discrepancies between simulation and actual measurement, and difficulty in reflecting the true mechanical response of robots. Summary of the Invention

[0005] This disclosure provides a robot performance prediction method, apparatus, device, and storage medium to achieve accurate prediction of robot performance.

[0006] Firstly, a method for predicting robot performance is provided, comprising: constructing a rigid body dynamics model of the robot based on structural parameter information and joint motion parameter information; simulating the rigid body dynamics model based on joint drive commands output by the motion control model to obtain mechanical and kinematic response parameter information output by the rigid body dynamics model; comparing the mechanical and kinematic response parameter information with the corresponding predicted parameter information output by the motion control model and the measured parameter information of the robot, and calibrating the rigid body dynamics model according to the consistency comparison result to obtain a calibrated model; making the key load-bearing components of the calibrated model flexible to obtain a rigid-flexible coupling dynamics model; inputting the measured robot motion data into the rigid-flexible coupling dynamics model to determine the dynamic stress result of the key load-bearing components; and performing a durability assessment of the key load-bearing components according to the dynamic stress result to obtain an assessment result.

[0007] This disclosure discloses an embodiment of a robot that constructs a rigid body dynamics model and uses joint drive commands output by a motion control model to simulate the rigid body dynamics model, obtaining mechanical and kinematic response parameters. These response parameters are then compared with the predicted parameters of the motion control model and the measured parameters of the robot. Based on the comparison results, the model is iteratively calibrated to obtain a calibrated model that highly matches the dynamic characteristics of the physical prototype. Furthermore, key load-bearing components are made flexible to construct a rigid-flexible coupled dynamic model, and the dynamic stress results of the key components are determined by combining measured motion data, thereby achieving durability assessment based on dynamic stress. This significantly improves the confidence level of the simulation model, enabling accurate identification of the dynamic stress distribution and potential failure risks of components in a virtual environment. This allows for the prediction of fatigue life of key robot components before the physical prototype is manufactured, effectively shortening the development cycle and reducing testing costs.

[0008] In one implementation, the rigid body dynamics model is simulated based on the joint drive commands output by the motion control model to obtain the mechanical and kinematic response parameter information output by the rigid body dynamics model. This includes: using the joint drive commands output by the motion control model as input to drive the rigid body dynamics model to perform simulation, obtaining simulation results containing the mechanical and kinematic response parameter information; or, using the joint drive commands output by the motion control model as input to drive the rigid body dynamics model to perform simulation, obtaining simulation results containing the mechanical and kinematic response parameter information; feeding back the response parameter information to the motion control model as the basis for adjusting the joint drive commands at the next moment; and completing collaborative simulation and outputting the final mechanical and kinematic response parameter information through real-time interactive iteration between the rigid body dynamics model and the motion control model.

[0009] In one of the above embodiments, the joint drive commands output by the motion control model are used as input to drive the rigid body dynamics model for simulation, directly yielding simulation results containing mechanical and kinematic response parameter information. Alternatively, in another embodiment, the drive commands output by the motion control model are input into the rigid body model, and the motion state of the rigid body model is fed back to the motion control model in real time, forming a bidirectional interactive iteration. This simulation mechanism enables the control algorithm to perceive the real response characteristics of the mechanical system in a virtual environment, optimize the control strategy in advance, and avoid repeated modifications caused by mismatch between control and structure in the later stages.

[0010] In one embodiment, the mechanical and kinematic response parameter information includes at least one of the following categories: joint load parameter information; contact mechanical parameter information; motion deviation parameter information; system energy parameter information; structural vibration parameter information; kinematic response parameter information; and the corresponding predicted parameter information synchronously output by the motion control model includes at least one of the following categories: expected joint load parameter information; expected contact mechanical parameter information; expected motion deviation parameter information; expected system energy parameter information; expected structural vibration parameter information; and expected kinematic response parameter information.

[0011] This implementation defines multiple parameters, including joint load, contact mechanics, motion deviation, system energy, structural vibration, and kinematic response, and maps them to the expected parameters of the motion control model. This multi-dimensional parameter system can provide unified data support for various aspects such as structural strength verification, control strategy evaluation, energy consumption analysis, and vibration suppression.

[0012] In one implementation, the mechanical and kinematic response parameter information is compared with the corresponding predicted parameter information output by the motion control model and the measured parameter information of the robot for consistency, and the rigid body dynamics model is calibrated based on the consistency comparison result to obtain a calibrated model. This includes: using the corresponding predicted parameter information output by the motion control model and the measured parameter information of the robot as reference parameter information, calculating the parameter deviation information between the mechanical and kinematic response parameter information and the reference parameter information; comparing the amplitude deviation rate and trend correlation coefficient corresponding to the parameter deviation information with corresponding consistency thresholds to generate a consistency comparison result; when the consistency comparison result indicates that the corresponding consistency threshold has not been reached, adjusting the physical environment parameter information of the rigid body dynamics model, re-executing the simulation, and obtaining updated mechanical and kinematic response parameter information; repeating the steps of parameter deviation information calculation, consistency comparison, and physical environment parameter adjustment until the consistency comparison result indicates that the corresponding consistency threshold has been reached, and using the finally adjusted rigid body dynamics model as the calibrated model.

[0013] Here, this implementation method uses the output of the motion control model and the actual measurement results of the robot as a benchmark to calculate the amplitude deviation rate and trend correlation coefficient of the mechanical and kinematic response parameters. After comparing them with the threshold, it determines whether to adjust the physical environment parameters and re-simulate, which can ensure the consistency between the dynamic characteristics of the model and the real system.

[0014] In one implementation, adjusting the physical environment parameter information of the rigid body dynamics model and re-executing the simulation includes: acquiring historical simulation data, which includes multiple sets of physical environment parameter information and their corresponding consistency comparison results; training a machine learning model based on the historical simulation data to obtain an automatic parameter optimization model; inputting the current parameter deviation information into the automatic parameter optimization model and outputting optimized physical environment parameter information; and re-executing the simulation based on the optimized physical environment parameter information.

[0015] This method trains a machine learning model by collecting historical simulation data and automatically recommends the optimal combination of physical environment parameters based on the current parameter deviation. This not only greatly improves debugging efficiency but also increases the upper limit of model accuracy.

[0016] In one embodiment, the key load-bearing components of the calibrated model are made flexible to obtain a rigid-flexible coupled dynamic model. Measured robot motion data is input into the rigid-flexible coupled dynamic model to determine the dynamic stress results of the key load-bearing components. This includes: identifying key structural components bearing dynamic loads based on the calibrated model as key load-bearing components; performing finite element mesh generation on the key load-bearing components and assigning corresponding material properties to generate a corresponding flexible body model; replacing the corresponding rigid components in the calibrated model with the flexible body model to construct the rigid-flexible coupled dynamic model; inputting the measured robot motion data into the rigid-flexible coupled dynamic model, performing time-domain simulation calculations, and outputting the dynamic stress results of the key load-bearing components in the time domain.

[0017] Since the rigidity assumption cannot obtain the deformation and stress information of the component in actual motion, by identifying key load-bearing components in the calibrated model, making them flexible, and inputting measured motion data for time-domain simulation, it is possible to capture the stress time history of the component under dynamic load, identify stress concentration areas and peak times, thereby improving the prediction accuracy.

[0018] In one embodiment, a durability assessment of the robot component is performed based on the dynamic stress results to obtain an assessment result, including: performing a durability assessment of the robot component based on the dynamic stress results and acquired material fatigue performance data to obtain an assessment result; or, collecting vibration load data at a specified test location on the robot, constructing a load characteristic spectrum based on the vibration load data, and generating an accelerated test spectrum equivalent to the load characteristic spectrum; and performing a durability assessment of the robot component based on the dynamic stress results and the accelerated test spectrum to obtain an assessment result.

[0019] One of the above implementation methods allows for direct comparison of fatigue life performance of different structural designs (such as L-shaped components and thighs) even without a physical prototype. The second implementation method utilizes accelerated testing data derived from vibration measurements taken by the robot in real-world scenarios (such as over speed bumps or potholes), which can reproduce specific damage experienced by the robot in actual use, thereby improving prediction accuracy.

[0020] In one implementation, vibration load data is collected at a designated test location on the robot. A load characteristic spectrum is constructed based on the vibration load data, and an equivalent accelerated test spectrum is generated. This includes: installing sensors at the designated test location on the robot and collecting acceleration signals under multiple preset test scenarios to obtain vibration load data; determining the fatigue damage spectrum and impact response spectrum for each test scenario based on the vibration load data; performing superposition processing on the fatigue damage spectra under each test scenario according to preset test specifications, and enveloping processing on the impact response spectra under each test scenario to obtain a load characteristic spectrum characterizing damage features under actual working conditions; the load characteristic spectrum includes the superimposed fatigue damage spectrum and the enveloped impact response spectrum; calculating the ultimate response spectrum based on the load characteristic spectrum, and performing envelope verification on the ultimate response spectrum using the enveloped impact response spectrum; when the enveloped impact response spectrum can envelop the ultimate response spectrum, converting the load characteristic spectrum into an equivalent accelerated test spectrum.

[0021] Because general testing standards may not match the specific working conditions of robots, laboratory test results may fail to reflect field performance. This implementation method collects vibration data from real-world scenarios, calculates fatigue damage and impact response spectra, and performs envelope verification after superposition and envelope processing to ensure that the generated acceleration spectrum does not lose key damage characteristics. Furthermore, generating an equivalent acceleration test spectrum significantly reduces test time while ensuring that the load is equivalent to the damage under actual working conditions, making the test results more authentic and reliable.

[0022] In one embodiment, a durability assessment of robot components is performed based on the dynamic stress results and the accelerated test spectrum to obtain an assessment result, including: extracting the stress time history of each key load-bearing component based on the dynamic stress results; performing rainflow counting processing on the stress time history to identify stress cycle information, wherein the stress cycle information includes the amplitude, mean, and number of occurrences of each type of stress cycle; determining the fatigue damage accumulation coefficient corresponding to the stress cycle information based on the accelerated test spectrum; determining the cumulative fatigue damage of each key load-bearing component based on the stress cycle information and the fatigue damage accumulation coefficient; and predicting the fatigue life of each key load-bearing component based on the cumulative fatigue damage and the material fatigue characteristic data of each key load-bearing component to obtain the assessment result.

[0023] This implementation extracts the stress time history from dynamic stress results, identifies the amplitude, mean, and number of stress cycles through rainflow counting, and calculates cumulative fatigue damage and predicts life by combining the damage coefficient determined by accelerated test spectrum and material fatigue curve, thereby achieving quantitative prediction and providing data support for subsequent structural optimization and maintenance strategies.

[0024] In one embodiment, the rigid body dynamics model is simulated based on the joint drive commands output by the motion control model to obtain the mechanical and kinematic response parameter information output by the rigid body dynamics model. This includes: obtaining the electromagnetic torque fluctuation parameter information of the robot motor and the frictional heat effect parameter information of the reducer, and establishing a multiphysics coupling model; integrating the multiphysics coupling model into the rigid body dynamics model, and taking into account the interference of electromagnetic force pulsation on joint torque and the material property changes caused by the temperature rise of the reducer in real time during the simulation process, and outputting the mechanical and kinematic response parameter information containing the multiphysics coupling effect.

[0025] Here, by integrating electromagnetic torque fluctuations and frictional heat effects into the rigid body dynamics model, torque disturbances and material property changes are taken into account in real time. This allows the simulation to predict failure modes that cannot be captured by purely mechanical models, such as structural reliability under high-temperature environments and fatigue damage caused by torque fluctuations.

[0026] In one embodiment, the method further includes: establishing a digital twin model of the robot, the digital twin model being built based on the calibrated model; collecting joint state data and acceleration data of the robot in real time through sensors, and uploading the joint state data and acceleration data to the digital twin model, so that the motion state of the digital twin model is synchronized with the robot in real time; based on the real-time state of the synchronized digital twin model, calling the calibrated model to perform online simulation calculations to obtain the real-time stress results of key load-bearing components under the current test scenario; generating optimized control parameters for reducing structural stress based on the real-time stress results; and sending the optimized control parameters to the robot's control system to adjust the robot's joint driving torque to reduce structural stress.

[0027] Here, a digital twin is built based on the calibrated model. The physical prototype's state is synchronized through real-time sensor data, and the stress of key components is calculated online, with optimized control parameters output in reverse. This combined virtual and physical control method allows the robot to actively optimize control by sensing the forces acting on it, maximizing operational performance while ensuring structural safety.

[0028] In one embodiment, the method further includes: acquiring real-time operating condition data of the robot during actual operation through a digital twin model, the operating condition data including load cycle count, load distribution, and cumulative working time; generating operational load history data based on the operating condition data; dynamically correcting the remaining fatigue life of each key load-bearing component using cumulative damage theory based on the operational load history data and the evaluation results, to obtain a dynamic life evaluation result; and generating maintenance reminder information and outputting a component maintenance plan containing suggested maintenance time and maintenance content when the remaining life of any key load-bearing component is lower than a preset maintenance threshold based on the dynamic life evaluation result.

[0029] Since conventional static lifespan predictions cannot reflect actual operating condition changes, maintenance timing is often inaccurate. This implementation method uses digital twins to collect real-time operating condition data and dynamically adjusts the remaining fatigue life of each key component. When the remaining lifespan falls below a threshold, it proactively issues a warning and outputs the maintenance window and maintenance requirements. This predictive maintenance approach avoids the resource waste of scheduled maintenance and the safety risks of post-failure repairs, thus balancing maintenance costs with robot performance and safety.

[0030] In summary, the robot performance prediction method, apparatus, device, and storage medium provided in this disclosure construct a robot rigid body dynamics model and perform simulation using a motion control model. This obtains the mechanical and kinematic response parameters of the rigid body dynamics model, compares them with motion control prediction data and measured data, and iteratively calibrates the model to obtain a high-precision calibrated model. Furthermore, the key components of the calibrated model are made more flexible, and the dynamic stress results are determined by combining measured motion data, ultimately achieving durability assessment. This method significantly improves simulation confidence through model collaboration and virtual-real calibration, enabling accurate prediction of component fatigue life in a virtual environment, thereby shortening the development cycle and reducing costs.

[0031] Specifically, firstly, by simulating the rigid body dynamics model using a motion control model, the joint drive command generation logic during the simulation process is made consistent with the motion control strategy of the real robot, improving the realism and credibility of the simulation results. Secondly, by comparing the mechanical and kinematic response parameters of the rigid body dynamics model with the predicted parameters output by the motion control model and the measured parameters of the robot, and iteratively calibrating the rigid body dynamics model based on the comparison results, the control logic of the simulation model is made highly consistent with the dynamic behavior of the real robot. This significantly reduces the deviation between simulation results and measured data, providing a high-precision model foundation for subsequent stress analysis and durability assessment. Based on the calibrated high-precision model, key components are made flexible to construct a rigid-flexible coupling model. This allows the simulation to calculate how the stress inside the component dynamically changes during actual movement, thereby accurately identifying stress concentration points and potential failure locations. Based on this, the durability assessment results are more consistent with the actual use conditions of the robot, improving the accuracy of fatigue life prediction. Furthermore, by using measured robot motion data as input for stress analysis, the determination of dynamic stress results is based on the robot's actual motion state, effectively combining simulation analysis with measured data and avoiding the separation between simulation and measurement. Finally, the durability of the key-bearing components is evaluated based on the dynamic stress results, thereby completing the virtual verification of robot performance and life prediction. This effectively reduces the number of iterations of the physical prototype, shortens the R&D cycle, and lowers R&D costs.

[0032] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the technical solutions of this disclosure.

[0033] To make the above-mentioned objects, features and advantages of this disclosure more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0034] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings used in the embodiments will be briefly described below. These drawings are incorporated in and constitute a part of this specification. They illustrate embodiments conforming to this disclosure and, together with the specification, serve to explain the technical solutions of this disclosure. It should be understood that the following drawings only show some embodiments of this disclosure and should not be considered as limiting the scope. Those skilled in the art can obtain other related drawings based on these drawings without creative effort.

[0035] Figure 1 A flowchart of a robot performance prediction method provided in this disclosure embodiment;

[0036] Figure 2A schematic diagram of a robot performance prediction process provided as an exemplary embodiment of this disclosure; Figure 3 This is a schematic diagram of a robot performance prediction device provided in an embodiment of the present disclosure; Figure 4 This is a schematic diagram of the structure of a device 400 provided in an embodiment of the present disclosure. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. The components of the embodiments of this disclosure described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this disclosure provided in the accompanying drawings is not intended to limit the scope of the claimed disclosure, but merely represents selected embodiments of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without inventive effort are within the scope of protection of this disclosure.

[0038] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0039] In this document, the term "and / or" merely describes a relationship, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0040] Furthermore, the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein.

[0041] The robot performance prediction method according to the embodiments of this disclosure will be described in further detail below.

[0042] This disclosure provides a method for predicting robot performance. The robot can be a legged robot; exemplarily, the legged robot can refer to a quadruped robot (such as a robot dog). Figure 1 As shown, the robot performance prediction method provided in this disclosure embodiment may include: S101: Construct a rigid body dynamics model of the robot based on its structural parameters and joint motion parameters.

[0043] Here, the robot's structural parameters include mass, inertia, and geometric dimensions, while its joint motion parameters include joint type, axis direction, and range of motion. Based on these structural and joint motion parameters, a multi-rigid-body dynamics model describing the robot's overall motion characteristics can be established, including defining rigid-body inertial parameters, constructing kinematic topology, and setting joint constraints and drives.

[0044] In practical implementation, the mass, center of mass position, moment of inertia, and geometric dimensions of each rigid body component can be extracted from the robot's computer-aided design (CAD) model or design drawings. The moment of inertia is typically defined relative to the center of mass coordinate system, forming an inertia tensor matrix. Based on the robot's mechanism design, the type of each joint (e.g., rotary joint, prismatic joint), the direction vector of the joint axis (representation in the local coordinate system), the range of motion of the joint (angular or displacement limits), and the initial position can be determined. Based on the robot's assembly relationships, the connection methods and kinematic tree structure between the rigid bodies can be determined, including the determination of the base (root rigid body), the parent-child relationships of each rigid body, and the force transmission paths.

[0045] Based on the above input parameters, a rigid body dynamics model can be constructed according to the following steps: First, define the rigid bodies and their inertial properties. For example, define the mass properties (mass value of each rigid body), center-of-mass position (coordinates of each rigid body's center of mass in its own coordinate system), and inertia properties (rotational inertia tensor of each rigid body relative to its center-of-mass coordinate system) for each component of the robot (including the base, body, legs, links, etc.). Then, establish the kinematic topology. Based on the connection relationships between the rigid bodies, construct a kinematic tree describing the overall structure of the robot: determine the base as the root node of the kinematic tree; establish the parent-child hierarchical relationship between the rigid bodies according to the order from the base to the end effector; define the predecessor (parent) and successor (child) rigid bodies connected to each joint, ensuring that the force transmission path is complete and free of redundant constraints. Finally, set the joint models and constraints. Each moving joint is parameterized, including joint type (setting the degree of freedom type of the joint according to the actual mechanism, including rotary joints, translational joints, ball joints, etc.), joint axis (defining the direction vector of the joint axis in the local coordinate system), joint constraints (setting the joint's range of motion, initial angle or initial position, and possible motion coupling relationships), and joint drive (setting the drive function according to the joint motion parameters (such as the time history of angle and angular velocity), which can be achieved by interpolating the measured motion data using spline curves as the drive input for the joint). Next, coordinate system transformation relationships are defined. Specifically, transformation matrices are established between each coordinate system, such as a tree transformation matrix, indicating a constant transformation from the parent rigid body coordinate system to the joint's predecessor coordinate system, determined by the robot's geometry and assembly relationships; another example is the joint transformation matrix, indicating the transformation from the joint's predecessor coordinate system to the joint's successor coordinate system, which dynamically changes with joint motion parameters (such as joint angles). Through the composite operation of the transformation matrices, the overall transformation relationship from the parent rigid body coordinate system to the child rigid body coordinate system can be obtained. Furthermore, external forces and initial conditions can be applied. For example, when setting up an external force field, you can apply gravitational acceleration and define possible contact forces, external loads, spring damping forces, etc.; when setting up initial conditions, you can set the initial position, initial velocity, and initial acceleration of each rigid body.

[0046] Based on the rigid body properties, topology, joint constraints, and driving information defined above, multi-rigid-body dynamic equations are automatically established. In practice, commercial multibody dynamics software or open-source simulation platforms can be used to automatically establish and solve the dynamic equations.

[0047] After completing the above steps, a rigid body dynamics model of the robot is obtained. This model can describe the robot's motion response under given drive and external load, and output mechanical and kinematic response parameters such as torque, velocity, acceleration of each joint, and contact forces between rigid bodies.

[0048] S102: Simulate the rigid body dynamics model based on the joint driving commands output by the motion control model, and obtain the mechanical and kinematic response parameter information output by the rigid body dynamics model.

[0049] Here, the motion control model is a mathematical model that simulates the motion control algorithm of a real robot. The rigid body dynamics model is simulated based on the output of the motion control model. The motion control model generates joint drive commands according to the control strategy; the rigid body dynamics model performs simulation calculations based on these joint drive commands and outputs mechanical and kinematic response parameters, including joint torques and contact forces.

[0050] In one implementation, the joint drive commands output by the motion control model can be directly used as input to drive the rigid body dynamics model to perform simulation, thereby obtaining simulation results containing the mechanical and kinematic response parameter information.

[0051] In the above implementation, a motion control model for generating joint drive commands is first constructed based on the actual control strategy parameters used in the robot motion control system. These control strategy parameters include, but are not limited to, controller gain parameters (such as proportional, integral, and derivative gains of a PID controller), impedance control parameters (such as target stiffness coefficient and target damping coefficient), kinematic constraint parameters (such as joint velocity limits and acceleration limits), and trajectory planning parameters. By introducing these realistic control strategy parameters, the constructed motion control model can simulate the decision-making logic and behavioral characteristics of the physical robot controller. Then, the joint drive commands output by the motion control model (such as desired joint torque, desired joint angle, or desired joint current) are used as input to drive the rigid body dynamics model for simulation calculations. Based on the received drive commands, the rigid body dynamics model calculates simulation results containing mechanical response (such as joint torque, contact force, and constraint reaction force) and kinematic response (such as joint displacement, velocity, acceleration, and angular velocity) parameters, taking into account physical factors such as gravity, inertia, joint friction, and contact force.

[0052] In another implementation, the motion control model and the rigid body dynamics model can be co-simulated. Specifically, the joint drive commands output by the motion control model are used as input to drive the rigid body dynamics model to perform simulation, obtaining simulation results containing mechanical and kinematic response parameter information; the response parameter information is fed back to the motion control model as the basis for the motion control model to adjust the joint drive commands at the next moment; through real-time interactive iteration between the rigid body dynamics model and the motion control model, the co-simulation is completed and the final mechanical and kinematic response parameter information is output.

[0053] In the above implementation, the output of the motion control model is used to simulate the rigid body dynamics model, and the response parameter information from the simulation results, such as the acquired kinematic response parameter information (reflecting joint motion state information), is fed back to the motion control model in real time as the basis for the motion control model to adjust the joint driving command at the next moment. Based on the deviation between the current actual motion state and the desired motion trajectory, the motion control model recalculates and outputs updated joint driving commands in conjunction with control strategy parameters, driving the rigid body dynamics model to perform simulation calculations for the next moment. Finally, through real-time interactive iteration between the rigid body dynamics model and the motion control model, a collaborative simulation process is formed. The above steps are repeated until the preset simulation duration ends, and the final mechanical and kinematic response parameter information is output for subsequent model calibration and performance analysis.

[0054] By using the above-mentioned collaborative simulation method, the joint drive command generation logic in the simulation process is kept consistent with the motion control strategy of the real robot, which can significantly improve the realism and credibility of the simulation results.

[0055] For example, the mechanical and kinematic response parameter information includes at least one of the following categories: joint load parameter information; contact mechanical parameter information; motion deviation parameter information; system energy parameter information; structural vibration parameter information; and kinematic response parameter information. The corresponding predicted parameter information synchronously output by the motion control model includes at least one of the following categories: expected joint load parameter information; expected contact mechanical parameter information; expected motion deviation parameter information; expected system energy parameter information; expected structural vibration parameter information; and expected kinematic response parameter information.

[0056] The joint load parameters described above reflect the loads borne by each joint during movement, including joint driving torque, joint axial force, joint radial force, and joint constraint reaction force. These parameters are used to evaluate the workload and structural strength of the joint drive system. The contact mechanics parameters described above reflect the interaction forces between the robot and its environment (such as the ground and obstacles), including wheel-ground contact force, foot contact force, contact stress distribution, contact pressure center location, and pressure cloud map of the contact area. These parameters are key inputs for analyzing the robot's walking stability, ground adaptability, and structural impact response. The motion deviation parameters described above reflect the difference between the robot's actual motion state and its desired motion state, including joint position error (the difference between the actual joint angle and the desired joint angle), joint velocity error (the difference between the actual joint angular velocity and the desired joint angular velocity), and end effector pose error. These parameters are used to evaluate the robot's motion control accuracy and trajectory tracking performance. The system energy parameters described above reflect the robot's energy consumption characteristics during movement, including the instantaneous power of each joint, the instantaneous power of the entire robot, cumulative energy consumption, and energy recovery efficiency. These parameters are used to evaluate the robot's energy efficiency and endurance. The structural vibration parameters described above reflect the vibration characteristics of key parts of the robot during movement, including vibration acceleration, velocity, displacement, and spectral distribution at specified measurement points (such as the hub motor side, the center of gravity of the robot body, and the end effector). These parameters are used to analyze the robot's dynamic stiffness, resonance characteristics, and vibration fatigue damage. The kinematic response parameters described above describe the geometric and temporal properties of each component of the robot during movement, including joint angles, joint angular velocities, joint angular accelerations, and the pose, velocity, and acceleration of the end effector. These parameters intuitively reflect the robot's motion state and trajectory characteristics, and are core indicators for evaluating the robot's kinematic performance (such as reachability and motion smoothness), as well as the basis for verifying whether the rigid body dynamics model accurately reproduces the target motion.

[0057] Corresponding to the aforementioned mechanical and kinematic response parameters, the corresponding predicted parameters output synchronously by the motion control model include: Desired joint load parameters such as the desired joint driving torque and desired joint axial force calculated by the motion control model according to the control strategy, serving as the predicted baseline values ​​for joint loads; Desired contact mechanics parameters, which are the contact mechanics baseline values ​​predicted by the motion control model based on the desired motion trajectory and environmental model, including the desired contact force and desired contact stress distribution; Desired motion deviation parameters, including the desired motion deviation values ​​under ideal conditions (usually zero or close to zero), serving as the evaluation baseline for motion accuracy; Desired system energy parameters, including the desired power consumption and desired energy consumption calculated based on the ideal model, serving as the baseline values ​​for energy efficiency evaluation; and Desired structural vibration parameters, including the desired vibration parameters under ideal rigid body motion (usually zero), serving as the evaluation baseline for vibration suppression effect. The aforementioned desired kinematic response parameters describe the ideal motion target set by the motion control model according to the control strategy, such as desired joint angles, desired joint angular velocities, desired joint angular accelerations, desired end-effector poses, and desired end-effector velocities and accelerations. These parameters serve as a reference for measuring the deviation between the actual motion state and the target, and are also a prerequisite for achieving accurate trajectory tracking and motion planning.

[0058] Based on the above simulation process, to further improve the simulation accuracy and make it closer to the actual mechanical response of the robot under high-intensity, long-term operation scenarios, this disclosure also provides a simulation method that includes multi-physics coupling effects. For example, the rigid body dynamics model is simulated based on the joint drive commands output by the motion control model, and the mechanical and kinematic response parameter information output by the rigid body dynamics model is obtained. This includes: obtaining the electromagnetic torque fluctuation parameter information of the robot motor and the frictional heat effect parameter information of the reducer, and establishing a multi-physics coupling model; integrating the multi-physics coupling model into the rigid body dynamics model, and taking into account in real time the interference of electromagnetic force pulsation on joint torque and the material property changes caused by the temperature rise of the reducer during the simulation process, outputting mechanical and kinematic response parameter information including multi-physics coupling effects.

[0059] Here, the electromagnetic torque fluctuation parameters can include the torque fluctuation amplitude, fluctuation frequency, and harmonic components of the motor under different speeds and load conditions, which can be obtained through finite element simulation or bench testing of the motor. The frictional heat effect parameters can include the functional relationship between the reducer's friction coefficient and temperature, heat generation rate, thermal conductivity, and the mapping relationship between material properties (such as elastic modulus and yield strength) and temperature, which can be obtained through reducer thermal characteristic tests or material handbooks. Based on the above parameters, an electromagnetic-thermal-mechanical multiphysics coupling model is established to describe the interaction between the electromagnetic field, temperature field, and structural mechanical field. This model can reflect the interference mechanism of electromagnetic force pulsation on joint torque, as well as the material property degradation law caused by the reducer's temperature rise. The multiphysics coupling model is integrated into the rigid body dynamics model, making it a component of the simulation.

[0060] In the above simulation process, the following coupling effects can be taken into account in real time: interference of electromagnetic force pulsation on joint torque: based on the current motor speed and load conditions, the electromagnetic torque fluctuation is dynamically calculated and superimposed on the joint drive command output by the motion control model; changes in material properties caused by the temperature rise of the reducer: based on the cumulative heat generation and heat conduction model of the reducer, material parameters such as friction coefficient and elastic modulus are updated in real time to make them change dynamically with temperature. Finally, the mechanical and kinematic response parameter information containing multi-physics coupling effects is output.

[0061] Through the enhanced simulation process described above, mechanical and kinematic response parameters that more closely resemble real physical processes are output, including joint torques affected by electromagnetic fluctuations, contact forces considering temperature rise effects, and dynamic stress results reflecting changes in material properties. These parameters provide more accurate inputs for subsequent model calibration, stress analysis, and durability assessment.

[0062] S103: Compare the mechanical and kinematic response parameter information with the corresponding predicted parameter information output by the motion control model and the measured parameter information of the robot, and calibrate the rigid body dynamic model according to the consistency comparison result to obtain the calibrated model (matching the dynamic characteristics of the physical prototype).

[0063] Here, the mechanical and kinematic response parameters output by the simulation are compared with the parameters predicted by the motion control model and the measured parameters of the robot. Based on the comparison results, the rigid body dynamics model is iteratively calibrated to obtain a calibrated model that matches the dynamic characteristics of the physical prototype.

[0064] Specifically, the mechanical and kinematic response parameters (such as actual joint torques and actual contact forces) output by the rigid body dynamics model in step S102 are compared and analyzed with the corresponding predicted parameters (such as expected joint torques and expected contact forces) output synchronously by the motion control model. Simultaneously, measured parameters (such as measured joint torques, measured contact forces, measured joint angles, and measured accelerations) collected by sensors under the same working conditions (e.g., under the same control output and operating environment) are introduced as another comparison benchmark. By quantifying the degree of deviation between the simulation results and the motion control predictions, and between the simulation results and the measured values, a comprehensive judgment is made as to whether the current rigid body dynamics model accurately reflects the dynamic behavior of the real robot. If any set of deviations exceeds the allowable range, the physical environment parameters in the model are adjusted, and the simulation is re-executed until the simulation results, predicted parameters, and robot measured parameters reach the preset consistency requirements, thereby completing the model calibration. This step calibrates the simulation model to have the same dynamic characteristics as the physical prototype, and by introducing measured data as a calibration benchmark, it ensures that the simulation model is not only aligned with the motion control logic, but also highly matched with the real dynamic characteristics of the physical prototype.

[0065] In one implementation, the mechanical and kinematic response parameter information is compared with the corresponding predicted parameter information output by the motion control model and the measured parameter information of the robot for consistency. Based on the consistency comparison result, the rigid body dynamics model is calibrated to obtain a calibrated model. This process may include: using the corresponding predicted parameter information output by the motion control model and the measured parameter information of the robot as reference parameter information, calculating the parameter deviation information between the mechanical and kinematic response parameter information and the reference parameter information; comparing the amplitude deviation rate and trend correlation coefficient corresponding to the parameter deviation information with corresponding consistency thresholds to generate consistency comparison results; when the consistency comparison result indicates that the corresponding consistency threshold has not been reached, adjusting the physical environment parameter information in the rigid body dynamics model, re-executing the simulation process, and obtaining updated mechanical and kinematic response parameter information; repeating the steps of parameter deviation information calculation, consistency comparison, and physical environment parameter adjustment until the consistency comparison result indicates that the corresponding consistency threshold has been reached, and using the finally adjusted rigid body dynamics model as the calibrated model.

[0066] Based on step S103 above, this implementation provides a specific method for performing consistency comparison and iterative calibration. For example, firstly, the corresponding predicted parameter information output by the motion control model and the robot's measured parameter information are used as reference parameter information, and the deviation between the mechanical and kinematic response parameter information and the reference parameter information is calculated. For example, the parameter deviation information includes, but is not limited to, amplitude deviation (reflecting differences in parameter magnitude), phase deviation (reflecting time lag or lead), and waveform deviation (reflecting the similarity of changing trends), etc. The above deviation calculations need to be performed separately for the simulation results and motion control predicted values, and for the simulation results and measured values. Secondly, the amplitude deviation rate and trend correlation coefficient corresponding to the parameter deviation information are compared with pre-set consistency thresholds, wherein different threshold standards can be set for the motion control predicted values ​​and the measured values. The amplitude deviation rate is used to quantify the closeness of parameter amplitudes, and the trend correlation coefficient (such as the Pearson correlation coefficient) is used to quantify the similarity of parameter changing trends. When the amplitude deviation rate and trend correlation coefficient between the simulation results and the motion control prediction values, and between the simulation results and the measured values, reach the corresponding thresholds, the consistency comparison is deemed passed; otherwise, it is deemed failed. Next, when the consistency comparison result indicates that the corresponding consistency threshold has not been reached, i.e., the consistency comparison has failed, the physical environment parameters in the rigid body dynamics model are adjusted. For example, the physical environment parameters include, but are not limited to, parameters affecting the system's dynamic response such as joint damping coefficient, joint stiffness coefficient, ground friction coefficient, and contact stiffness coefficient. Finally, based on the adjusted physical environment parameters, the simulation process of step S102 is re-executed to obtain updated mechanical and kinematic response parameters. The steps of parameter deviation calculation, consistency comparison, and physical environment parameter adjustment are repeated until the consistency comparison result indicates that the deviation between the simulation results and the motion control prediction values ​​and measured values ​​reaches the corresponding consistency threshold. The finally adjusted rigid body dynamics model is used as the calibrated model for subsequent stress analysis and durability assessment.

[0067] Through the above iterative calibration process, the control logic of the simulation model is made to closely match the dynamic behavior of the real robot, which can significantly reduce the deviation between the simulation results and the measured data.

[0068] Based on the above embodiments, to further improve the efficiency and intelligence of parameter adjustment, this disclosure also provides a machine learning-based automatic parameter optimization method. In one embodiment, adjusting the physical environment parameter information of the rigid body dynamics model and re-executing the simulation may include: acquiring historical simulation data, which includes multiple sets of physical environment parameter information and their corresponding consistency comparison results; training a machine learning model based on the historical simulation data to obtain an automatic parameter optimization model; inputting the current parameter deviation information into the automatic parameter optimization model and outputting optimized physical environment parameter information; and re-executing the simulation process based on the optimized physical environment parameter information.

[0069] For example, adjusting the physical environment parameters of the rigid body dynamics model and re-executing the simulation can include: First, acquiring historical simulation data, which includes multiple sets of physical environment parameter information and their corresponding consistency comparison results. Each set of data records the combination of physical environment parameters used in a simulation (such as joint damping coefficient, ground friction coefficient, etc.), and the parameter deviation information or consistency compliance status calculated through consistency comparison under that parameter combination. Then, based on the historical simulation data, a machine learning model is trained to obtain an automatic parameter optimization model. The machine learning model can employ algorithms such as gradient boosting trees, neural networks, support vector regression, or Gaussian process regression to learn the mapping relationship between physical environment parameters and simulation accuracy, establishing an intelligent model capable of predicting the optimal parameter combination. Next, the parameter deviation information calculated in the current iteration is input into the trained automatic parameter optimization model, which automatically outputs the optimized physical environment parameter information. Compared with manual trial and error, this method can comprehensively consider the coupling relationship between multiple parameters and quickly locate the parameter combination that optimizes the simulation accuracy. Finally, based on the optimized physical environment parameter information, the simulation process is re-executed to obtain updated mechanical and kinematic response parameter information, and consistency comparison is continued until the calibration requirements are met.

[0070] By introducing machine learning models, the parameter tuning process, which originally relied on human experience, is transformed into a data-driven automatic optimization process, which significantly improves calibration efficiency and the optimization of parameter combination selection.

[0071] S104: Make the key load-bearing components of the calibrated model flexible to obtain a rigid-flexible coupling dynamic model. Input the measured robot motion data into the rigid-flexible coupling dynamic model to determine the dynamic stress results of the key load-bearing components.

[0072] Here, after obtaining a calibrated model that matches the dynamic characteristics of the physical prototype, it is necessary to further evaluate the stress conditions and distribution of key structural components during actual robot operation. Because these components bear complex dynamic loads during robot movement, their stress state changes over time, and traditional static analysis cannot accurately reflect their true stress characteristics. Therefore, this step involves making the key load-bearing components in the calibrated model more flexible, constructing a rigid-flexible coupled dynamic model, and using measured robot motion data (such as joint angles, joint angular velocities, and joint torques) as input to drive the rigid-flexible coupled model to perform time-domain simulation calculations. The output is the dynamic stress results of the key load-bearing components in the time domain, including the stress-time history of each node, stress cloud map, and stress amplitude in stress concentration areas. In subsequent durability assessments, these dynamic stress results can be used to identify stress concentration areas, assess fatigue damage risk, and provide quantitative basis for component optimization design. This rigid-flexible coupled analysis fully considers the influence of component elastic deformation on dynamic response, making the stress calculation results closer to the actual stress state of the component, and providing more accurate input data for subsequent fatigue life prediction.

[0073] In one embodiment, the key load-bearing components of the calibrated model are made flexible to obtain a rigid-flexible coupled dynamic model. Measured robot motion data is input into the rigid-flexible coupled dynamic model to determine the dynamic stress results of the key load-bearing components. This process may include: identifying key structural components bearing dynamic loads based on the calibrated model as key load-bearing components; performing finite element mesh generation on the key load-bearing components and assigning corresponding material properties to generate a corresponding flexible body model; replacing the corresponding rigid components in the calibrated model with the flexible body model to construct the rigid-flexible coupled dynamic model; inputting the measured robot motion data into the rigid-flexible coupled dynamic model, performing time-domain simulation calculations, and outputting the dynamic stress results of the key load-bearing components in the time domain.

[0074] Based on step S104 above, this embodiment provides a specific implementation method for determining the dynamic stress results of key load-bearing components based on a rigid-flexible coupling dynamic model. First, based on the calibrated model, key structural components that bear the main dynamic loads during robot movement are identified and designated as key load-bearing components. These key load-bearing components include, but are not limited to, components in the robot prone to stress concentration or fatigue failure, such as L-shaped parts, thighs, calves, wheel hub connectors, joint connectors, and base mounting seats. These components are typically located on the main load-bearing path, and their stress state has a decisive impact on the structural integrity and reliability of the robot. As one implementation method, rigid body dynamics simulation can be performed based on the calibrated model to output the load spectrum of each connection point of the robot under typical operating conditions. The load spectrum includes joint moments, connection point forces, and contact forces. Based on the load spectrum, the main load-bearing path of the robot is determined. The main load-bearing path is a continuous force transmission channel from the point of application to the robot body. Structural components located on the main load-bearing path and subjected to alternating dynamic loads are identified as key load-bearing components.

[0075] Then, finite element meshes are generated for the identified key load-bearing components, and corresponding material properties (such as elastic modulus, Poisson's ratio, density, yield strength, etc.) are assigned to them to generate corresponding flexible body models. Unlike rigid body models, flexible body models can describe the deformation, strain, and stress distribution of components under stress, and are the basis for dynamic stress analysis. The density of the finite element mesh should be balanced according to the requirements of computational accuracy and efficiency, and a local refinement strategy can be adopted in stress concentration areas. Next, the generated flexible body model replaces the corresponding rigid components in the calibrated model to construct a rigid-flexible coupling dynamic model. In this model, the key load-bearing components exist in the form of flexible bodies, capable of elastic deformation and stress response; while other non-critical components remain in the form of rigid bodies to ensure computational efficiency. By defining connection constraints (such as fixed constraints, hinge constraints, etc.) between the flexible and rigid bodies, the force transmission path is ensured to be consistent with the original model. Finally, measured robot motion data (such as joint angle time history, joint torque time history, etc.) are input into the rigid-flexible coupling dynamic model to perform time-domain simulation calculations. During the simulation, the flexible body model calculates its deformation and stress state in real time based on the dynamic load it receives, and outputs the dynamic stress results of key load-bearing components over time, including the stress-time history of each node, stress cloud map, stress amplitude of stress concentration areas, and other information.

[0076] Through the above rigid-flexible coupling stress analysis, the dynamic stress distribution of key load-bearing components under actual motion can be accurately obtained while taking into account the flexibility of the components, providing high-precision input data for subsequent fatigue damage calculation and life assessment.

[0077] S105: Based on the dynamic stress results, perform a durability assessment on the key load-bearing components to obtain the assessment results.

[0078] In one embodiment, the durability of the robot component can be evaluated based on the dynamic stress results and the acquired material fatigue performance data to obtain the evaluation results.

[0079] Here, after obtaining the dynamic stress results of key load-bearing components, it is necessary to further evaluate the durability performance of these components under actual working conditions. Based on the dynamic stress results output in step S104 (such as stress-time history, stress amplitude distribution, etc.), combined with the fatigue performance data of the material, the fatigue damage accumulation theory is used to predict the life of key load-bearing components. For example, the stress-time history of each key load-bearing component can be extracted from the dynamic stress results first; then, rainflow counting processing can be performed on the stress-time history to identify the amplitude, mean, and number of stress cycles; finally, based on the material's fatigue characteristic curve (SN curve) and Mainner's linear cumulative damage rule, the cumulative fatigue damage of each component is calculated and its fatigue life is predicted.

[0080] In another implementation, based on the principle of equal damage, the original load spectrum can be compressed into a shorter accelerated test spectrum to efficiently reproduce the fatigue damage effect of actual working conditions under laboratory conditions. Specifically, vibration load data at a specified test location on the robot can be collected, a load characteristic spectrum can be constructed based on the vibration load data, and an accelerated test spectrum equivalent to the load characteristic spectrum can be generated; based on the dynamic stress results and the accelerated test spectrum, the durability of the robot component can be evaluated to obtain the evaluation results.

[0081] For example, sensors can first be installed at designated test locations on the robot to collect vibration acceleration data under typical scenarios. The collected data is then processed and analyzed to construct a load characteristic spectrum that characterizes damage under actual working conditions. Based on the principle of equal damage, an accelerated test spectrum equivalent to the original load spectrum is generated for subsequent durability assessment. This approach allows the durability assessment results to better reflect the robot's actual usage conditions.

[0082] Here, collecting vibration load data at a designated test location on the robot, constructing a load characteristic spectrum based on the vibration load data, and generating an acceleration test spectrum equivalent to the load characteristic spectrum can include: installing sensors at the designated test location on the robot, collecting acceleration signals under multiple preset test scenarios to obtain vibration load data; determining the fatigue damage spectrum and impact response spectrum for each test scenario based on the vibration load data; performing superposition processing on the fatigue damage spectrum for each test scenario according to preset test specifications, and performing envelope processing on the impact response spectrum for each test scenario to obtain a load characteristic spectrum used to characterize the damage characteristics under actual working conditions; the load characteristic spectrum includes the superimposed fatigue damage spectrum and the enveloped impact response spectrum; calculating the ultimate response spectrum based on the load characteristic spectrum, and performing envelope verification on the ultimate response spectrum using the enveloped impact response spectrum; when the enveloped impact response spectrum can enclose the ultimate response spectrum, converting the load characteristic spectrum into an equivalent acceleration test spectrum.

[0083] Based on step S105 above, the above implementation provides a specific method for generating an equivalent acceleration test spectrum based on measured vibration data. For example, firstly, sensors are installed at designated test locations on the robot to collect acceleration signals under multiple preset test scenarios, obtaining vibration load data. The designated test locations include, but are not limited to, key measuring points such as the hub motor side, joint connections, and the center of gravity of the robot body; the multiple preset test scenarios include combinations of different travel speeds, different load conditions, and different road surface types. By covering typical operating conditions in actual robot operation, the representativeness and comprehensiveness of the collected vibration data are ensured. Secondly, the collected vibration load data is preprocessed to calculate the fatigue damage spectrum (FDS) and shock response spectrum (SRS) for each test scenario. The fatigue damage spectrum describes the cumulative fatigue damage generated when the structure is subjected to vibration loads at different frequencies; the shock response spectrum describes the relationship between the maximum response acceleration and the natural frequency of a single-degree-of-freedom system under shock excitation. These two types of spectra characterize the damage features of the vibration load from different perspectives. Subsequently, according to the preset test specifications, the fatigue damage spectra under each test scenario are superimposed to obtain a superimposed fatigue damage spectrum that represents the cumulative damage effect under all working conditions. Simultaneously, the impact response spectra under each test scenario are enveloped to obtain an enveloped impact response spectrum that covers the most severe impact conditions across all working conditions. The superimposed fatigue damage spectrum and the enveloped impact response spectrum are used together as a load characteristic spectrum to characterize the damage features under actual working conditions. Then, the extreme response spectrum (ERS) is calculated based on the load characteristic spectrum, and the enveloped impact response spectrum is used to perform envelope verification on the extreme response spectrum. The extreme response spectrum is a reference spectrum calculated from the accelerated test spectrum to assess whether the test conditions are too severe or insufficient. The purpose of envelope verification is to ensure that the generated accelerated test spectrum can reproduce the most severe impact conditions in actual working conditions, avoiding undertesting due to excessively weak test conditions or overtesting due to excessively strong test conditions. Finally, when the enveloped impact response spectrum can enclose the limiting response spectrum, it indicates that the current load characteristic spectrum meets the envelope requirement, and it is converted into an accelerated test spectrum equivalent to the original load spectrum. If the envelope verification fails, the test time parameters need to be adjusted, and the superposition, envelope, and verification steps need to be repeated until the envelope requirement is met.

[0084] Through the above steps, an equivalent accelerated test spectrum based on real working conditions and verified by envelope is generated, which can provide accurate and efficient test input for subsequent durability assessment.

[0085] When evaluating the durability of robot components based on the dynamic stress results and the accelerated testing spectrum, for example, the stress-time history of key load-bearing components can be extracted from the dynamic stress results and decomposed into a series of stress cycles (including amplitude, mean, and number of cycles). Then, the fatigue damage accumulation coefficient corresponding to each stress cycle is determined based on the accelerated testing spectrum; next, the cumulative fatigue damage of each component is calculated. Finally, combined with the fatigue characteristic curve of the material, the remaining fatigue life of each component is predicted, yielding the evaluation result. This evaluation result can be used to identify potential failure risks, guide structural optimization design, and provide a quantitative basis for subsequent maintenance decisions, achieving virtual verification of robot durability performance before physical prototype manufacturing.

[0086] Based on this, the present disclosure also provides a specific implementation method for durability assessment based on dynamic stress results and accelerated test spectra. In one embodiment, durability assessment of robot components is performed based on the dynamic stress results and the accelerated test spectra to obtain assessment results, including: extracting the stress time history of each key load-bearing component based on the dynamic stress results; performing rainflow counting processing on the stress time history to identify stress cycle information, wherein the stress cycle information includes the amplitude, mean, and number of occurrences of each type of stress cycle; determining the fatigue damage accumulation coefficient corresponding to the stress cycle information based on the accelerated test spectra; determining the cumulative fatigue damage of each key load-bearing component based on the stress cycle information and the fatigue damage accumulation coefficient; and predicting the fatigue life of each key load-bearing component based on the cumulative fatigue damage and the material fatigue characteristic curves of each key load-bearing component to obtain the assessment results.

[0087] Here, based on the dynamic stress results obtained in step S104, stress-time history data of each key load-bearing component in the time domain are extracted. This data describes the stress variation of the component during robot movement over time. Rainflow counting is performed on the stress-time history to identify stress cycle information. Here, the rainflow counting method is a standard counting method that decomposes a random stress-time history into complete stress cycles. By identifying closed hysteresis loops in the stress-time curve, the amplitude, mean, and number of occurrences of each stress cycle are extracted. This cycle information reflects the fatigue load spectrum characteristics of the component under actual working conditions.

[0088] Based on the accelerated test spectrum generated in step S105, the fatigue damage accumulation coefficient corresponding to the stress cycle information is determined. This fatigue damage accumulation coefficient can be calculated based on the principle of equal damage, using the acceleration coefficient and material fatigue characteristic curve parameters, and is used to reflect the proportion of fatigue damage caused by each stress cycle under accelerated test conditions relative to the original working condition. According to the identified stress cycle information and the corresponding fatigue damage accumulation coefficient, the cumulative fatigue damage of each key load-bearing component is calculated using cumulative damage theory (such as Miner's linear cumulative damage rule). Specifically, for each type of stress cycle, its cycle number is divided by the allowable cycle number at that stress level (obtained from the material fatigue characteristic curve) to obtain the damage component caused by that type of cycle; the damage components of all cycles are summed to obtain the total cumulative fatigue damage. Finally, based on the calculated cumulative fatigue damage and combined with the material fatigue characteristic curves of each key load-bearing component, the fatigue life of each component is predicted. For example, by judging whether the cumulative fatigue damage value reaches the preset failure threshold, it is determined whether the component has experienced fatigue failure; and based on the current cumulative damage value and the operating time, the remaining fatigue life can be calculated to obtain the durability assessment result.

[0089] Through the above steps, a quantitative assessment from dynamic stress analysis to fatigue life prediction is achieved, providing a scientific basis for robot reliability design and maintenance decisions.

[0090] Based on the durability assessment results obtained above, to further realize the health management of the robot throughout its entire life cycle, this disclosure provides a dynamic life assessment and maintenance early warning method based on a digital twin model. A digital twin model can be considered an enhanced virtual entity that extends the functions of real-time data interaction, state synchronization, and reverse control on a calibrated model. In one embodiment, the method further includes: collecting real-time operating condition data of the robot during actual operation through the digital twin model, the operating condition data including load cycle count, load distribution, and cumulative working time; generating operational load history data based on the operating condition data; dynamically correcting the remaining fatigue life of each key load-bearing component using cumulative damage theory based on the operational load history data and the assessment results, obtaining a dynamic life assessment result; and generating a maintenance reminder message and outputting a component maintenance plan containing suggested maintenance time and content when the remaining life of any key load-bearing component is lower than a preset maintenance threshold based on the dynamic life assessment result.

[0091] For example, a digital twin model is used to collect real-time operational data of the robot during actual operation. This operational data includes, but is not limited to, the number of load cycles (e.g., the number of times the robot crosses speed bumps daily, the number of emergency stops), load distribution (e.g., the ratio of time spent operating under no-load and with load), and cumulative working time. This data reflects the robot's actual usage intensity and operating environment, forming the basis for dynamically correcting the lifespan assessment. Based on the real-time collected operational data, the robot's operational load history data is statistically generated. This history data records the various loads the robot experiences during actual use and their frequency of occurrence, supplementing and correcting the initial design conditions. Based on the operational load history data and the aforementioned durability assessment results, the cumulative damage theory is used to dynamically correct the remaining fatigue life of each key load-bearing component. Unlike the static initial lifespan assessment, dynamic correction considers the cumulative load effect of the robot during actual use, continuously updating the remaining lifespan prediction value as operating time increases, resulting in a dynamic lifespan assessment result. For example, when the robot is detected frequently crossing speed bumps, the fatigue damage accumulation rate of the corresponding components will accelerate, and the remaining lifespan will be shortened accordingly. Based on the dynamic lifespan assessment results, the remaining lifespan status of each key load-bearing component is monitored in real time. When the remaining lifespan of any critical load-bearing component falls below a preset maintenance threshold, a maintenance reminder is automatically generated. This maintenance threshold can be preset based on the component's safety importance and maintenance strategy, such as a remaining lifespan of less than 100 hours or less than 10% of the design lifespan. Finally, based on the maintenance reminder, a component maintenance plan containing suggested maintenance times and content is output. For example, specific maintenance suggestions such as "Left rear wheel hub connector has a remaining lifespan of 95 hours; inspection is recommended after 80 hours" or "Thigh component has accumulated damage of 0.85; replacement is recommended at the next maintenance" are output, providing decision support for maintenance personnel.

[0092] Through the aforementioned dynamic life assessment and maintenance early warning mechanism, the maintenance plan can be dynamically adjusted according to the actual usage of the robot, which avoids the waste of resources caused by premature maintenance and prevents the safety risks caused by late maintenance, thus significantly improving the availability and operation and maintenance economy of the robot.

[0093] Based on the aforementioned performance prediction method, this disclosure also provides a control method based on a digital twin model for dynamically optimizing control parameters to reduce structural stress during robot operation. Exemplarily, the method further includes: establishing a digital twin model of the robot, the digital twin model being built based on the calibrated model (the digital twin model can be considered an enhanced virtual entity that extends the calibrated model with real-time data interaction, state synchronization, and reverse control functions); collecting joint state data and acceleration data of the robot in real time through sensors, and uploading the joint state data and acceleration data to the digital twin model, so that the motion state of the digital twin model is synchronized with the robot in real time; based on the real-time state of the synchronized digital twin model, calling the calibrated model to perform online simulation calculations to obtain the real-time stress results of key load-bearing components under the current test scenario; generating optimized control parameters for reducing structural stress based on the real-time stress results; and sending the optimized control parameters to the robot's control system to adjust the robot's joint driving torque to reduce structural stress.

[0094] In practical implementation, a digital twin model of the robot is established based on the calibrated model obtained in the aforementioned steps. This digital twin model uses the calibrated model as the core of mechanical calculations and integrates a real-time data interaction interface, enabling it to synchronize with the physical robot in real time. Unlike the calibrated model used offline, the digital twin model can continuously receive real-time state data from the physical robot and output optimized control commands. Joint state data (such as joint angles, joint angular velocities, and joint torques) and acceleration data (such as fuselage acceleration and wheel hub acceleration) are collected in real time by sensors installed on the robot and uploaded to the digital twin model. The digital twin model updates its own motion state based on the received real-time data, ensuring consistency with the current state of the physical robot, achieving virtual-real synchronization. Based on the synchronized real-time state of the digital twin model, the calibrated model in its core is invoked for online simulation calculations. Unlike offline simulation, online simulation uses the current real-time state as the initial condition and predicts the system response in the short time domain, thereby obtaining the real-time stress results of key load-bearing components in the current test scenario. Based on the real-time stress results, optimized control parameters are generated with the goal of reducing structural stress. The optimized control parameters can include adjusted joint drive torques, modified impedance control parameters, or replanned desired trajectories, used to reduce stress levels in critical components by altering the control strategy. Finally, the generated optimized control parameters are sent to the physical robot's control system to adjust the joint drive torques in real time, thereby reducing structural stress in critical load-bearing components. This achieves real-time interaction and collaborative optimization between the simulation model and the physical prototype. This method enables dynamic adjustment of the control strategy during robot operation, effectively preventing overload and fatigue damage, and improving the robot's safety and reliability under complex working conditions.

[0095] like Figure 2 The diagram shown is a schematic representation of robot performance prediction provided in an embodiment of this disclosure. A rigid body dynamics model of the robot is constructed based on structural parameters output from the robot's structural design model, such as the 3D model, mass, center of mass, and moment of inertia, as well as joint motion parameters such as joint type, axial direction, and range of motion.

[0096] The robot's motion is controlled by a motion control model, which outputs joint drive commands (such as joint rotation angles) and predicted parameter information (such as joint drive torque, wheel-to-ground contact force, and IMU acceleration). The joint drive commands output by the motion control model are then input into the robot's rigid body dynamics model for simulation. The rigid body dynamics model outputs mechanical and kinematic response parameters (such as actual joint torque and actual contact force). Simultaneously, sensors installed on the robot collect measured parameters at designated locations, including vibration and mechanical parameters (such as acceleration at the front leg and rear wheel hub positions, and IMU acceleration).

[0097] The mechanical and kinematic response parameters output from the rigid body dynamics model are compared with the corresponding predicted parameters (such as expected joint torques and expected contact forces) synchronously output from the motion control model, as well as the measured parameters of the robot, to complete the calibration of the rigid body dynamics model. The key load-bearing components of the calibrated model, such as the L-shaped joint, thigh, lower leg, and wheel hub, are made flexible to obtain a rigid-flexible coupled dynamic model. The dynamic stress results of the key load-bearing components in the rigid-flexible coupled dynamic model are determined by combining the measured robot motion data. Finally, based on the dynamic stress results, the durability of the key load-bearing components is evaluated, and the evaluation results are obtained. If the performance evaluation fails, the model is returned to the structural design model for redesign and simulation of the robot's structural parameters until the performance evaluation is passed.

[0098] like Figure 3 As shown, this disclosure provides a robot performance prediction device 300, comprising: Model building module 31 is used to build a rigid body dynamics model of the robot based on the robot's structural parameters and joint motion parameters; Simulation module 32 is used to simulate the rigid body dynamics model based on the joint drive commands of the motion control model, and obtain the mechanical and motion response parameter information output by the rigid body dynamics model; The calibration module 33 is used to compare the mechanical and kinematic response parameter information with the corresponding predicted parameter information output by the motion control model and the measured parameter information of the robot, and to calibrate the rigid body dynamics model according to the consistency comparison result to obtain the calibrated model. The stress analysis module 34 is used to make the key load-bearing components of the calibrated model flexible to obtain a rigid-flexible coupling dynamic model. The measured robot motion data is input into the rigid-flexible coupling dynamic model to determine the dynamic stress results of the key load-bearing components. The evaluation module 35 is used to perform a durability evaluation on the robot component based on the dynamic stress results, and obtain the evaluation results.

[0099] In one embodiment, the simulation module 32 is specifically used to: take the joint drive command output by the motion control model as input to drive the rigid body dynamics model to perform simulation, and obtain simulation results containing the mechanical and kinematic response parameter information; or, take the joint drive command output by the motion control model as input to drive the rigid body dynamics model to perform simulation, and obtain simulation results containing the mechanical and kinematic response parameter information; feed back the response parameter information to the motion control model as the basis for the motion control model to adjust the joint drive command at the next moment; and complete the collaborative simulation and output the final mechanical and kinematic response parameter information through real-time interactive iteration between the rigid body dynamics model and the motion control model.

[0100] In one embodiment, the mechanical and kinematic response parameter information includes at least one of the following categories: joint load parameter information; contact mechanical parameter information; motion deviation parameter information; system energy parameter information; structural vibration parameter information; kinematic response parameter information; and the corresponding predicted parameter information synchronously output by the motion control model includes at least one of the following categories: expected joint load parameter information; expected contact mechanical parameter information; expected motion deviation parameter information; expected system energy parameter information; expected structural vibration parameter information; and expected kinematic response parameter information.

[0101] In one embodiment, the calibration module 33 is specifically used to: use the corresponding predicted parameter information output by the motion control model and the measured parameter information of the robot as reference parameter information, calculate the parameter deviation information between the mechanical and kinematic response parameter information and the reference parameter information; compare the amplitude deviation rate and trend correlation coefficient corresponding to the parameter deviation information with the corresponding consistency thresholds respectively, and generate a consistency comparison result; when the consistency comparison result indicates that the corresponding consistency threshold has not been reached, adjust the physical environment parameter information of the rigid body dynamics model, re-execute the simulation, and obtain updated mechanical and kinematic response parameter information; repeat the steps of parameter deviation information calculation, consistency comparison, and physical environment parameter adjustment until the consistency comparison result indicates that the corresponding consistency threshold has been reached, and use the finally adjusted rigid body dynamics model as the calibrated model.

[0102] In one implementation, when the calibration module 33 adjusts the physical environment parameter information of the rigid body dynamics model and re-invokes the simulation module 32 to execute the simulation, it is specifically used to: acquire historical simulation data, which includes multiple sets of physical environment parameter information and their corresponding consistency comparison results; train a machine learning model based on the historical simulation data to obtain an automatic parameter optimization model; input the current parameter deviation information into the automatic parameter optimization model and output the optimized physical environment parameter information; and re-invoke the simulation module 32 to execute the simulation based on the optimized physical environment parameter information.

[0103] In one embodiment, the stress analysis module 34 is specifically used for: identifying key structural components bearing dynamic loads based on the calibrated model, as key load-bearing components; performing finite element mesh generation on the key load-bearing components and assigning corresponding material properties to generate a corresponding flexible body model; replacing the corresponding rigid component in the calibrated model with the flexible body model to construct a rigid-flexible coupling dynamic model; inputting the measured robot motion data into the rigid-flexible coupling dynamic model, performing time-domain simulation calculations, and outputting the dynamic stress results of the key load-bearing components in the time domain.

[0104] In one embodiment, the evaluation module 35 is specifically used to: perform a durability evaluation on the robot component based on the dynamic stress results and the acquired material fatigue performance data, and obtain an evaluation result; or, collect vibration load data at a specified test location of the robot, construct a load characteristic spectrum based on the vibration load data, and generate an accelerated test spectrum equivalent to the load characteristic spectrum; and perform a durability evaluation on the robot component based on the dynamic stress results and the accelerated test spectrum, and obtain an evaluation result.

[0105] In one embodiment, the evaluation module 35 is specifically used for: installing sensors at designated test locations on the robot, acquiring acceleration signals under multiple preset test scenarios, and obtaining vibration load data; determining the fatigue damage spectrum and impact response spectrum for each test scenario based on the vibration load data; performing superposition processing on the fatigue damage spectrum for each test scenario according to preset test specifications, and performing envelope processing on the impact response spectrum for each test scenario to obtain a load feature spectrum for characterizing the damage characteristics under actual working conditions; the load feature spectrum includes the superimposed fatigue damage spectrum and the enveloped impact response spectrum; calculating the ultimate response spectrum based on the load feature spectrum, and performing envelope verification on the ultimate response spectrum using the enveloped impact response spectrum; when the enveloped impact response spectrum can enclose the ultimate response spectrum, converting the load feature spectrum into an equivalent acceleration test spectrum.

[0106] In one embodiment, the simulation module 32 is specifically used to: acquire electromagnetic torque fluctuation parameter information of the robot motor and frictional heat effect parameter information of the reducer, and establish a multi-physics coupling model; integrate the multi-physics coupling model into the rigid body dynamics model, and take into account the interference of electromagnetic force pulsation on joint torque and the material property changes caused by the temperature rise of the reducer in real time during the simulation process, and output mechanical and kinematic response parameter information containing the multi-physics coupling effect.

[0107] In one embodiment, the robot performance prediction device 300 further includes: The collaborative control execution module 36 is used to establish a digital twin model of the robot, which is based on the calibrated model. It collects joint state data and acceleration data of the robot in real time through sensors and uploads the joint state data and acceleration data to the digital twin model, synchronizing the motion state of the digital twin model with that of the robot in real time. Based on the real-time state of the synchronized digital twin model, it calls the calibrated model to perform online simulation calculations to obtain the real-time stress results of key load-bearing components under the current test scenario. Based on the real-time stress results, it generates optimized control parameters to reduce structural stress. The optimized control parameters are then sent to the robot's control system to adjust the robot's joint driving torque to reduce structural stress.

[0108] In one embodiment, the evaluation module 35 is specifically used for: extracting the stress time history of each key load-bearing component based on the dynamic stress results; performing rainflow counting processing on the stress time history to identify stress cycle information, the stress cycle information including the amplitude, mean, and number of occurrences of each type of stress cycle; determining the fatigue damage accumulation coefficient corresponding to the stress cycle information based on the accelerated test spectrum; determining the cumulative fatigue damage of each key load-bearing component according to the stress cycle information and the fatigue damage accumulation coefficient; and predicting the fatigue life of each key load-bearing component based on the cumulative fatigue damage and the material fatigue characteristic curve of each key load-bearing component, thereby obtaining the evaluation result.

[0109] In one embodiment, the evaluation module 35 is further configured to: collect real-time operating condition data of the robot during actual operation through a digital twin model, the operating condition data including load cycle count, load distribution and cumulative working time; generate operating load history data based on the operating condition data; dynamically correct the remaining fatigue life of each key load-bearing component using cumulative damage theory based on the operating load history data and the evaluation results, to obtain a dynamic life evaluation result; and generate maintenance reminder information and output a component maintenance plan containing suggested maintenance time and maintenance content when the remaining life of any key load-bearing component is lower than a preset maintenance threshold based on the dynamic life evaluation result.

[0110] For detailed implementation of each of the above modules, please refer to the description of the aforementioned methods; it will not be repeated here.

[0111] Reference Figure 4 The diagram shown is a schematic representation of the structure of a device 400 according to an exemplary embodiment of this disclosure. The device 400 can be a computer device, a control chip, etc., and can be deployed on a server, a terminal, a robot, a remote control device, etc. It may include a processor 410, a memory 420, and a bus 430. The memory 420 is used to store execution instructions and includes main memory 421 and external memory 422. The main memory 421, also called internal memory, is used to temporarily store computational data in the processor 410 and data exchanged with external memory 422 such as a hard disk. The processor 410 exchanges data with the external memory 422 through the main memory 421.

[0112] In this embodiment, the memory 420 is specifically used to store application code executing the scheme of this disclosure, and its execution is controlled by the processor 410. That is, when the electronic device 400 is running, the processor 410 communicates with the memory 420 through the bus 430, or the processor 410 communicates with the memory 420 through other means, so that the processor 410 executes the application code stored in the memory 420, thereby executing the steps of the robot performance prediction method described in any of the foregoing embodiments. The memory 420 may be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The processor 410 may be an integrated circuit chip with signal processing capabilities. The aforementioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor.

[0113] This disclosure also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the robot performance prediction method described in any of the above embodiments. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. Available media can be magnetic media, such as hard disks, floppy disks, and magnetic tapes; optical media, such as DVD-ROM, DVD-RAM, DVD-RW, DVD+RW, CD-ROM, CD-RW, CD-RW, and MO (magneto-optical) storage media; and semiconductor storage media, such as flash memory, EEPROM, Dynamic Random Access Memory (DRAM), and Static Random Access Memory (SRAM).

[0114] The computer program can be written in various computer programming languages, including but not limited to C, C++, Python, and custom messages and services under the ROS framework. When the computer program is executed by the processor, it implements the various steps of the robot performance prediction method in the embodiments of this disclosure.

[0115] This disclosure also provides a computer program product storing a computer program. When run by a processor, the computer program executes the steps of the robot performance prediction method provided in any of the above embodiments of this disclosure. For details, please refer to the above method embodiments, which will not be repeated here. The computer program product can be implemented using hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied as a computer storage medium, which can be a volatile or non-volatile computer-readable storage medium. In another optional embodiment, the computer program product is specifically embodied as a software product, such as a software development kit (SDK), etc.

[0116] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices and apparatuses described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. In the several embodiments provided in this disclosure, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0117] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, the functional units in the various embodiments of this disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0118] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0119] Finally, it should be noted that the above-described embodiments are merely specific implementations of this disclosure, used to illustrate the technical solutions of this disclosure, and not to limit it. The protection scope of this disclosure is not limited thereto. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this disclosure; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this disclosure, and should all be covered within the protection scope of this disclosure. Therefore, the protection scope of this disclosure should be determined by the protection scope of the claims.

Claims

1. A method for predicting robot performance, characterized in that, include: A rigid body dynamics model of the robot is constructed based on the robot's structural parameters and joint motion parameters. The rigid body dynamics model is simulated based on the joint driving commands output by the motion control model to obtain the mechanical and kinematic response parameter information output by the rigid body dynamics model. The mechanical and kinematic response parameter information is compared with the corresponding predicted parameter information output by the motion control model and the measured parameter information of the robot. The rigid body dynamics model is calibrated according to the consistency comparison results to obtain the calibrated model. The key load-bearing components of the calibrated model are made flexible to obtain a rigid-flexible coupling dynamic model. The measured robot motion data is input into the rigid-flexible coupling dynamic model to determine the dynamic stress results of the key load-bearing components. Based on the dynamic stress results, a durability assessment is performed on the key load-bearing components to obtain the assessment results.

2. The method according to claim 1, characterized in that, The rigid body dynamics model is simulated based on the joint drive commands output by the motion control model to obtain the mechanical and kinematic response parameter information output by the rigid body dynamics model, including: The joint drive commands output by the motion control model are used as input to drive the rigid body dynamics model to perform simulation, and simulation results containing the mechanical and kinematic response parameter information are obtained. Alternatively, the joint drive commands output by the motion control model can be used as input to drive the rigid body dynamics model to perform simulation, obtaining simulation results containing the mechanical and kinematic response parameter information; the response parameter information can be fed back to the motion control model as the basis for the motion control model to adjust the joint drive commands at the next moment; through real-time interactive iteration between the rigid body dynamics model and the motion control model, co-simulation can be completed and the final mechanical and kinematic response parameter information can be output.

3. The method according to claim 1, characterized in that, The mechanical and kinematic response parameter information includes at least one of the following categories: joint load parameter information; Contact mechanics parameters; kinematic deviation parameters; system energy parameters; structural vibration parameters; kinematic response parameters; The corresponding predicted parameter information synchronously output by the motion control model includes at least one of the following categories: expected joint load parameter information; expected contact mechanics parameter information; expected motion deviation parameter information; expected system energy parameter information; expected structural vibration parameter information; and expected kinematic response parameter information.

4. The method according to claim 1, characterized in that, The mechanical and kinematic response parameter information is compared with the corresponding predicted parameter information output by the motion control model and the measured parameter information of the robot. Based on the consistency comparison results, the rigid body dynamics model is calibrated to obtain a calibrated model, including: Using the corresponding predicted parameter information output by the motion control model and the robot's measured parameter information as reference parameter information, the parameter deviation information between the mechanical and kinematic response parameter information and the reference parameter information is calculated; The amplitude deviation rate and trend correlation coefficient corresponding to the parameter deviation information are compared with the corresponding consistency thresholds to generate consistency comparison results. When the consistency comparison result indicates that the corresponding consistency threshold has not been reached, the physical environment parameter information of the rigid body dynamics model is adjusted, the simulation is re-executed, and updated mechanical and kinematic response parameter information is obtained. Repeat the steps of calculating the parameter deviation information, comparing consistency, and adjusting the physical environment parameters until the consistency comparison result indicates that the corresponding consistency threshold is reached, and use the finally adjusted rigid body dynamics model as the calibrated model.

5. The method according to claim 4, characterized in that, Adjusting the physical environment parameters of the rigid body dynamics model and re-running the simulation includes: Acquire historical simulation data, which includes multiple sets of physical environment parameter information and their corresponding consistency comparison results; Based on the historical simulation data, a machine learning model is trained to obtain an automatic parameter optimization model. Input the current parameter deviation information into the parameter automatic optimization model, and output the optimized physical environment parameter information; Based on the optimized physical environment parameter information, the simulation is re-executed.

6. The method according to claim 1, characterized in that, The key load-bearing components of the calibrated model are made flexible to obtain a rigid-flexible coupled dynamic model. Measured robot motion data is input into this model to determine the dynamic stress results of the key load-bearing components, including: Based on the calibrated model, key structural components that bear dynamic loads are identified as key load-bearing components. The key load-bearing components are meshed using finite element methods, and corresponding material properties are assigned to generate the corresponding flexible body model. Replace the corresponding rigid component in the calibrated model with the flexible body model to construct a rigid-flexible coupled dynamic model. The measured robot motion data is input into the rigid-flexible coupling dynamic model, time-domain simulation calculation is performed, and the dynamic stress results of the key load-bearing components in the time domain are output.

7. The method according to claim 1, characterized in that, Based on the dynamic stress results, a durability assessment is performed on the robot components to obtain the assessment results, including: Based on the dynamic stress results and the obtained material fatigue performance data, the durability of the robot component is evaluated, and the evaluation results are obtained. Alternatively, vibration load data at a designated test location on the robot can be collected, a load characteristic spectrum can be constructed based on the vibration load data, and an accelerated test spectrum equivalent to the load characteristic spectrum can be generated; based on the dynamic stress results and the accelerated test spectrum, the durability of the robot components can be evaluated to obtain the evaluation results.

8. The method according to claim 7, characterized in that, Collect vibration load data at a designated test location on the robot, construct a load characteristic spectrum based on the vibration load data, and generate an accelerated test spectrum equivalent to the load characteristic spectrum, including: Sensors are installed at designated test locations on the robot to collect acceleration signals under multiple preset test scenarios and obtain vibration load data. Based on the vibration load data, the fatigue damage spectrum and impact response spectrum under each test scenario are determined; According to the preset test specifications, the fatigue damage spectrum under each test scenario is superimposed, and the impact response spectrum under each test scenario is enveloped to obtain a load feature spectrum for characterizing the damage characteristics under actual working conditions; the load feature spectrum includes the superimposed fatigue damage spectrum and the enveloped impact response spectrum. The ultimate response spectrum is calculated based on the load characteristic spectrum, and the ultimate response spectrum is enveloped and verified using the enveloped impact response spectrum. When the enclosed impact response spectrum can enclose the limiting response spectrum, the load characteristic spectrum is converted into an equivalent accelerated test spectrum.

9. The method according to claim 7 or 8, characterized in that, Based on the dynamic stress results and the accelerated test spectrum, the robot components are subjected to a durability assessment, and the assessment results are obtained, including: Based on the dynamic stress results, the stress time history of each key load-bearing component is extracted; The stress time history is processed by rainflow counting to identify stress cycle information, which includes the amplitude, mean, and number of times the type of cycle occurs for each stress cycle. Based on the accelerated test spectrum, the fatigue damage accumulation coefficient corresponding to the stress cycle information is determined; Based on the stress cycle information and the fatigue damage accumulation coefficient, the cumulative fatigue damage of each key load-bearing component is determined; Based on the accumulated fatigue damage, and combined with the material fatigue performance data of each key load-bearing component, the fatigue life of each key load-bearing component is predicted, and the evaluation results are obtained.

10. The method according to claim 1, characterized in that, The rigid body dynamics model is simulated based on the joint drive commands output by the motion control model to obtain the mechanical and kinematic response parameter information output by the rigid body dynamics model, including: Obtain electromagnetic torque fluctuation parameters of the robot motor and frictional heat effect parameters of the reducer, and establish a multi-physics coupling model; The multiphysics coupling model is integrated into the rigid body dynamics model. During the simulation, the interference of electromagnetic force pulsation on joint torque and the material property changes caused by the temperature rise of the reducer are taken into account in real time. The mechanical and kinematic response parameter information containing the multiphysics coupling effect is output.

11. The method according to claim 1, characterized in that, The method further includes: A digital twin model of the robot is established, the digital twin model being built based on the calibrated model; The robot's joint status data and acceleration data are collected in real time by sensors, and the joint status data and acceleration data are uploaded to the digital twin model so that the motion state of the digital twin model is synchronized with the robot in real time. Based on the real-time state of the synchronized digital twin model, the calibrated model is called to perform online simulation calculations to obtain the real-time stress results of key load-bearing components under the current test scenario. Based on the real-time stress results, optimized control parameters for reducing structural stress are generated; The optimized control parameters are sent to the robot's control system to adjust the robot's joint drive torque to reduce structural stress.

12. The method according to claim 1, characterized in that, The method further includes: The robot's operating condition data is collected in real time through a digital twin model. The operating condition data includes the number of load cycles, load distribution, and cumulative working time. Based on the operating condition data, the operating load history data is statistically generated. Based on the operational load history data and the evaluation results, the remaining fatigue life of each key load-bearing component is dynamically corrected using the cumulative damage theory to obtain the dynamic life evaluation results. Based on the dynamic life assessment results, when the remaining life of any critical load-bearing component is lower than the preset maintenance threshold, a maintenance reminder message is generated, and a component maintenance plan containing suggested maintenance time and maintenance content is output.

13. A robot performance prediction device, characterized in that, include: The model building module is used to build a rigid body dynamics model of the robot based on its structural parameters and joint motion parameters. The simulation module is used to simulate the rigid body dynamics model based on the joint driving commands of the motion control model, and to obtain the mechanical and motion response parameter information output by the rigid body dynamics model. The calibration module is used to compare the mechanical and kinematic response parameter information with the corresponding predicted parameter information output by the motion control model and the measured parameter information of the robot, and to calibrate the rigid body dynamics model according to the consistency comparison result to obtain the calibrated model. The stress analysis module is used to make the key load-bearing components of the calibrated model more flexible to obtain a rigid-flexible coupling dynamic model. The measured robot motion data is input into the rigid-flexible coupling dynamic model to determine the dynamic stress results of the key load-bearing components. The evaluation module is used to perform a durability evaluation on the robot components based on the dynamic stress results, and obtain the evaluation results.

14. A device, characterized in that, It includes a processor and a memory, wherein the memory stores computer instructions, and the processor is used to execute the computer instructions to perform the steps of the robot performance prediction method according to any one of claims 1 to 12.

15. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the robot performance prediction method as described in any one of claims 1 to 12.

Images